prozak / rtk Goto Github PK

Implement the way to add user-specified 'import' clauses to the generated haskell code.

make test-grammar should regenerate files only if needed. Currently it doing so always. This improvement is needed for debugging of generated Lexer and Parser files.

Comment like /* ... */ should be supported. Nested multiline comments should be supported as well

In case when main rule could be empty this would lead to grammar ambiguity

Add support for SYB to the grammar data structures. This would allow to remove boilerplate code when traversing and rewriting grammar data structures.

Reaction on empty file should be adequate and not rtk.exe: Parse error[]

Identify why Alex file compilation in some cases takes too long. For now it java.pg and haskell.pg

It adds unneeded overhead during alex and happy work

Add normal support for 'star' to the grammar

More informative output when there is a lexer error. Now there is the following one:

rtk: lexical error

Fix 'addStartRule' from Grammar module and find a place for it in the normalization process.

Problem arose when added QQ vars. We cannot add vars to all rules that are lifted. Solution should be investigated.

Please, refer to java.pg

If the grammar file is:

grammar 'tapl1';
tapl1 = t;

then the following error message is printed:
rtk.exe: Maybe.fromJust: Nothing

Rules that have ',' * are not working

Happy syntax file generation implementation for the given grammar

subj needs interference with GenY, GenX and GenAST modules.

Now just rule with the type that is the same as rule can be in the list.
This is because during Normalization for the following rule:
*

Proxy rules should be added to the type of the .
But information on rule type is unavailable during the Normalization.
For now type named is used.

In order to reference one lexical rule from another one macro should be generated in Alex. Currently <TODO: macro ref here> is generated.

And automate it somehow for the makefile's test-* target

Somewhy ImportsOpt is not added for quasiquotation tokens. Needs investigation

Make Haskell grammar working with rtk

Errors like "rtk.exe: Wrong clause IRegExpLit " Arglist "" are non-informative. At least context (rule name) should be specified. It would be really convenient for big grammars.

After parsing file we receiving term that could not be matched against main grammar terminal. This rule should not be generated. It should be enough to add all QQ stuff in root non-terminal.

There should be a way to get primitive value from a terminal.
For example:

Name = id;

There should be either a functions, so one can write:

getId (n :: Name),

or user-defined constructor name:

let (IdName id) = (n :: Name)

See GenY module.

Add support for nested lexems. So for example multiline comments could be nested:

/* /* */ */

We still need to generate '-sub' term kinds for terms with loops in order to get list or to build term with list

Problem is that when QQ vars added some token needed that match the following rules:
[a-zA-Z][a-zA-Z_]*

There is already exists one in test-grammar.pg. So I cannot just add new one because they will overlap during lexer.
So, lexical rules should be normalized and possibility should be added to find token rule.

Add support for multi-line tokens, since generated now alex basic wrapped doesn't support multi-line tokens.